Public Versus Authority: How Different Targets of Algorithm‐Assisted Decision Making Impact on Citizens' Trust?

ABSTRACT Amid the rapid evolution of digital governance, algorithm‐assisted decision making (ADM) has transcended its functional role in public affairs management to become a significant carrier of administrative power. Based on principal‐agent theory, we propose a “public‐algorithms‐authority” triadic governance framework. This study examines in high‐stakes governance domains, how the algorithmic targets (i.e., the public via Internet opinion regulation versus supervising authority targets through dynamic judicial review) influence citizens' trust in ADM. We conducted a survey experiment in Shanghai, recruiting 1207 participants, to compare two empirical cases: ADM for “Internet public opinion regulation” and “dynamic judicial supervision”. The findings reveal that citizens exhibit higher trust in ADM is aimed to supervise authority rather than the public. Furthermore, citizens with greater algorithmic experience are more willing to trust ADM. The study identifies algorithmic black‐box and algorithmic flaws as mediators between algorithmic targets and citizens' trust in ADM. These insights provide practical policy recommendations for fostering public confidence in ADM.